AI Literacy / Training faculty
How do you train faculty on AI?
Not with a mandatory workshop. The institutions making progress started by taking the objections seriously.
Effective faculty AI training is short, self-paced, discipline-relevant, and voluntary at the start. Roughly seventy minutes of foundational content gets most faculty from uncertain to conversant. What does not work is a single mandatory in-service day, a tools demonstration, or training delivered by someone with no teaching experience.
The harder problem is not delivery. It is that a significant share of your faculty have principled objections to AI, and a program that treats those objections as resistance to be overcome will lose the people whose judgment you most need.
Why the in-service day does not work
The default institutional response is a session at fall convocation. Ninety minutes, whole faculty, an outside speaker, some slides about large language models, and a demonstration of a chatbot writing a lesson plan.
Three things go wrong. The room contains faculty who use these tools daily and faculty who have never opened one, so the content is simultaneously too basic and too advanced. The demonstration format teaches a product rather than a concept, and the product will have changed by spring. And the mandatory frame guarantees that the most skeptical people in the room experience the session as advocacy, which hardens the position it was meant to soften.
The faculty who most need the training are the ones least likely to attend a voluntary session and least likely to engage in a mandatory one. That is the actual design problem.
The faculty whose judgment you most need are the ones least persuaded by enthusiasm.
What the content has to cover
Concepts rather than tools, in an order that builds.
| Layer | What faculty learn | Why it matters in a classroom |
|---|---|---|
| Foundation | How these systems actually work | Without this, every downstream conversation about integrity, accuracy, or appropriate use is guesswork. It also explains why the tools fail in the specific ways they do. |
| Limits | Where output is unreliable | Fabricated citations, confident errors, and uneven subject accuracy are predictable rather than random. Faculty who understand why can design around them. |
| Integrity | Assignment design, not detection | Detection is unreliable and adversarial. Assessment redesign is durable. This is the highest-value session for most faculty. |
| Discipline | What it means in this field | The stakes in nursing, welding, English composition, and accounting are genuinely different. Generic training cannot address this and faculty know it. |
| Judgment | When not to use it | A program that never says no has no credibility. Naming the cases where these tools are inappropriate is what earns trust from skeptics. |
Reaching the skeptics
Faculty objections to AI are not uniform, and treating them as a single attitude is the most common design error. In our sessions they sort into four distinct positions, each of which needs a different response.
- The labor objectionThat AI is a cost-reduction project aimed at teaching positions. This one is often correct at other institutions, and the only useful response is your own institution being explicit about what it will and will not do.
- The pedagogical objectionThat struggling with a hard text is the learning, and a tool that removes the struggle removes the education. This is a serious position held by serious teachers and it deserves engagement, not a workaround.
- The ethical objectionTraining data provenance, labor conditions in data annotation, and environmental cost. Dismissing these forfeits credibility with exactly the faculty who read carefully.
- The fatigue objectionThat this is the fourth transformative technology they have been asked to adopt in a decade. Usually the easiest to address, because it is a request for evidence rather than enthusiasm.
A program that acknowledges all four, and concedes the parts that are true, gets further with skeptical faculty than one built on adoption metrics. It also produces better institutional decisions, because these objections identify real risks.
Structure that works
What we have seen hold up across community college and university deployments.
- Self-paced and short - under ten minutes per module
- Voluntary first term, expected thereafter
- Concept-led, so it survives product changes
- Discipline breakouts after the shared foundation
- Facilitated by people who have taught
- An explicit section on inappropriate use
- Department-level follow-up, not one central event
Our AI Foundations program is built on this structure: eight modules, roughly seventy minutes total, self-paced, no jargon. The first module is free and open, which matters more than it sounds - it lets a skeptical faculty member evaluate the program privately before committing to it, and it lets a dean send a link rather than schedule a meeting.
Institutions typically pair it with facilitated discipline sessions, where the shared vocabulary from the modules makes a ninety-minute departmental conversation productive instead of definitional.
For cabinets and senates specifically, the same content runs as a briefing rather than a course, because those groups need enough understanding to govern rather than to teach.
Who does this work
Former provosts, presidents, and system executives who have run professional development inside participatory governance.
How long does faculty AI training take?
About seventy minutes of foundational content is enough to move most faculty from uncertain to conversant, delivered as short self-paced modules rather than one session. Discipline-specific application usually needs a further ninety-minute facilitated conversation per department, which is where the real work happens.
Should faculty AI training be mandatory?
Voluntary for the first term, then expected. Mandating it before there is any peer credibility guarantees that the most skeptical faculty arrive already opposed. Once respected colleagues in a department have been through it and say it was worth the time, an expectation becomes reasonable.
What about faculty who object to AI on principle?
Take the objection seriously and separate its four forms: labor, pedagogical, ethical, and fatigue. Each needs a different response, and several are substantively correct. A program that concedes what is true gets much further with these faculty, and their scrutiny improves institutional decisions.
Should we train faculty on specific AI tools?
Teach concepts first, tools second and lightly. Product interfaces change within a term, so tool-specific training expires quickly, while an understanding of how these systems work and where they fail stays useful. Tool demonstrations belong in optional follow-up sessions.
Does AI training help with academic integrity?
Yes, but through assignment design rather than detection. Faculty who understand what these systems do well and poorly can build assessments that are difficult to complete without doing the underlying work. Detection tools are unreliable enough that no integrity strategy should depend on them.
Send your faculty the first module.
It is free, open, and takes about ten minutes. No form and no sign-up, which is what makes it something a dean can share in an email.
Related